If we have multiple unwritten extents within a single page, we fail to
tell the I/o completion construction handlers we need a new handle for the
second and subsequent blocks in the page. While we still issue the I/O
correctly, we do not have the correct ranges recorded in the ioend
structures and hence when we go to convert the unwritten extents we screw
it up.
Make sure we start a new ioend every time the mapping changes so that we
convert the correct ranges on I/O completion.
SGI-PV: 964647
SGI-Modid: xfs-linux-melb:xfs-kern:28797a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
With the per-cpu superblock counters, batch updates are no longer atomic
across the entire batch of changes. This is not an issue if each
individual change in the batch is applied atomically. Unfortunately, free
block count changes are not applied atomically, and they are applied in a
manner guaranteed to cause problems.
Essentially, the free block count reservation that the transaction took
initially is returned to the in core counters before a second delta takes
away what is used. because these two operations are not atomic, we can
race with another thread that can use the returned transaction reservation
before the transaction takes the space away again and we can then get
ENOSPC being reported in a spot where we don't have an ENOSPC condition,
nor should we ever see one there.
Fix it up by rolling the two deltas into the one so it can be applied
safely (i.e. atomically) to the incore counters.
SGI-PV: 964465
SGI-Modid: xfs-linux-melb:xfs-kern:28796a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Currently we do not wait on extent conversion to occur, and hence we can
return to userspace from a synchronous direct I/O write without having
completed all the actions in the write. Hence a read after the write may
see zeroes (unwritten extent) rather than the data that was written.
Block the I/O completion by triggering a synchronous workqueue flush to
ensure that the conversion has occurred before we return to userspace.
SGI-PV: 964092
SGI-Modid: xfs-linux-melb:xfs-kern:28775a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
When processing multiple extent maps, xfs_bmapi needs to keep track of the
extent behind the one it is currently working on to be able to trim extent
ranges correctly. Failing to update the previous pointer can result in
corrupted extent lists in memory and this will result in panics or assert
failures.
Update the previous pointer correctly when we move to the next extent to
process.
SGI-PV: 965631
SGI-Modid: xfs-linux-melb:xfs-kern:28773a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Vlad Apostolov <vapo@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
When we have a couple of hundred transactions on the fly at once, they all
typically modify the on disk superblock in some way.
create/unclink/mkdir/rmdir modify inode counts, allocation/freeing modify
free block counts.
When these counts are modified in a transaction, they must eventually lock
the superblock buffer and apply the mods. The buffer then remains locked
until the transaction is committed into the incore log buffer. The result
of this is that with enough transactions on the fly the incore superblock
buffer becomes a bottleneck.
The result of contention on the incore superblock buffer is that
transaction rates fall - the more pressure that is put on the superblock
buffer, the slower things go.
The key to removing the contention is to not require the superblock fields
in question to be locked. We do that by not marking the superblock dirty
in the transaction. IOWs, we modify the incore superblock but do not
modify the cached superblock buffer. In short, we do not log superblock
modifications to critical fields in the superblock on every transaction.
In fact we only do it just before we write the superblock to disk every
sync period or just before unmount.
This creates an interesting problem - if we don't log or write out the
fields in every transaction, then how do the values get recovered after a
crash? the answer is simple - we keep enough duplicate, logged information
in other structures that we can reconstruct the correct count after log
recovery has been performed.
It is the AGF and AGI structures that contain the duplicate information;
after recovery, we walk every AGI and AGF and sum their individual
counters to get the correct value, and we do a transaction into the log to
correct them. An optimisation of this is that if we have a clean unmount
record, we know the value in the superblock is correct, so we can avoid
the summation walk under normal conditions and so mount/recovery times do
not change under normal operation.
One wrinkle that was discovered during development was that the blocks
used in the freespace btrees are never accounted for in the AGF counters.
This was once a valid optimisation to make; when the filesystem is full,
the free space btrees are empty and consume no space. Hence when it
matters, the "accounting" is correct. But that means the when we do the
AGF summations, we would not have a correct count and xfs_check would
complain. Hence a new counter was added to track the number of blocks used
by the free space btrees. This is an *on-disk format change*.
As a result of this, lazy superblock counters are a mkfs option and at the
moment on linux there is no way to convert an old filesystem. This is
possible - xfs_db can be used to twiddle the right bits and then
xfs_repair will do the format conversion for you. Similarly, you can
convert backwards as well. At some point we'll add functionality to
xfs_admin to do the bit twiddling easily....
SGI-PV: 964999
SGI-Modid: xfs-linux-melb:xfs-kern:28652a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
If hole punching at EOF is done as two steps (i.e. truncate then extend)
the file is in a transient state between the two steps where an
application can see the incorrect file size. Punching a hole to EOF needs
to be treated in teh same way as all other hole punching cases so that the
file size is never seen to change.
SGI-PV: 962012
SGI-Modid: xfs-linux-melb:xfs-kern:28641a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Vlad Apostolov <vapo@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
When setting the length of the iclogbuf to write out we should just be
changing the desired byte count rather completely reassociating the buffer
memory with the buffer. Reassociating the buffer memory changes the
apparent length of the buffer and hence when we free the buffer, we don't
free all the vmap()d space we originally allocated.
SGI-PV: 964983
SGI-Modid: xfs-linux-melb:xfs-kern:28640a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Don't reference the log buffer after running the callbacks as the callback
can trigger the log buffers to be freed during unmount.
SGI-PV: 964545
SGI-Modid: xfs-linux-melb:xfs-kern:28567a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Recent fixes to the filesystem freezing code introduced a vn_iowait call
in the middle of the sync code. Unfortunately, at the point where this
call was added we are holding the ilock. The ilock is needed by I/O
completion for unwritten extent conversion and now updating the file size.
Hence I/o cannot complete if we hold the ilock while waiting for I/O
completion.
Fix up the bug and clean the code up around it.
SGI-PV: 963674
SGI-Modid: xfs-linux-melb:xfs-kern:28566a
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Tim Shimmin <tes@sgi.com>
When growing a filesystem we don't check to see if the new size overflows
the page cache index range, so we can do silly things like grow a
filesystem page 16TB on a 32bit. Check new filesystem sizes against the
limits the kernel can support.
SGI-PV: 957886
SGI-Modid: xfs-linux-melb:xfs-kern:28563a
Signed-Off-By: Nathan Scott <nscott@aconex.com>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
Many block drivers (aoe, iscsi) really want refcountable pages in bios,
which is what almost everyone send down. XFS unfortunately has a few
places where it sends down buffers that may come from kmalloc, which
breaks them.
Fix the places that use kmalloc()d buffers.
SGI-PV: 964546
SGI-Modid: xfs-linux-melb:xfs-kern:28562a
Signed-Off-By: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <dgc@sgi.com>
Signed-off-by: Tim Shimmin <tes@sgi.com>
* git://git.linux-nfs.org/pub/linux/nfs-2.6: (122 commits)
sunrpc: drop BKL around wrap and unwrap
NFSv4: Make sure unlock is really an unlock when cancelling a lock
NLM: fix source address of callback to client
SUNRPC client: add interface for binding to a local address
SUNRPC server: record the destination address of a request
SUNRPC: cleanup transport creation argument passing
NFSv4: Make the NFS state model work with the nosharedcache mount option
NFS: Error when mounting the same filesystem with different options
NFS: Add the mount option "nosharecache"
NFS: Add support for mounting NFSv4 file systems with string options
NFS: Add final pieces to support in-kernel mount option parsing
NFS: Introduce generic mount client API
NFS: Add enums and match tables for mount option parsing
NFS: Improve debugging output in NFS in-kernel mount client
NFS: Clean up in-kernel NFS mount
NFS: Remake nfsroot_mount as a permanent part of NFS client
SUNRPC: Add a convenient default for the hostname when calling rpc_create()
SUNRPC: Rename rpcb_getport to be consistent with new rpcb_getport_sync name
SUNRPC: Rename rpcb_getport_external routine
SUNRPC: Allow rpcbind requests to be interrupted by a signal.
...
When nfsd was transitioned to use splice instead of sendfile() for data
transfers, a line setting the page index was lost. Restore it, so that
nfsd is functional when that path is used.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
If the output actor doesn't transfer the full amount of data, we will
increment ppos too much. Two related bugs in there:
- We need to break out and return actor() retval if it is shorted than
what we spliced into the pipe.
- Adjust ppos only according to actor() return.
Also fix loop problem in generic_file_splice_read(), it should not keep
going when data has already been transferred.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Revalidate read/write permissions for splice(2) and vmslice(2), in case
security policy has changed since the files were opened.
Acked-by: Stephen Smalley <sds@tycho.nsa.gov>
Signed-off-by: James Morris <jmorris@namei.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Well, first of all, I don't want to change so many files either.
What I do:
Adding a new parameter "struct bin_attribute *" in the
.read/.write methods for the sysfs binary attributes.
In fact, only the four lines change in fs/sysfs/bin.c and
include/linux/sysfs.h do the real work.
But I have to update all the files that use binary attributes
to make them compatible with the new .read and .write methods.
I'm not sure if I missed any. :(
Why I do this:
For a sysfs attribute, we can get a pointer pointing to the
struct attribute in the .show/.store method,
while we can't do this for the binary attributes.
I don't know why this is different, but this does make it not
so handy to use the binary attributes as the regular ones.
So I think this patch is reasonable. :)
Who benefits from it:
The patch that exposes ACPI tables in sysfs
requires such an improvement.
All the table binary attributes share the same .read method.
Parameter "struct bin_attribute *" is used to get
the table signature and instance number which are used to
distinguish different ACPI table binary attributes.
Without this parameter, we need to offer different .read methods
for different ACPI table binary attributes.
This is impossible as there are various ACPI tables on different
platforms, and we don't know what they are until they are loaded.
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch makes dentries and inodes for sysfs directories
reclaimable.
* sysfs_notify() is modified to walk sysfs_dirent tree instead of
dentry tree.
* sysfs_update_file() and sysfs_chmod_file() use sysfs_get_dentry() to
grab the victim dentry.
* sysfs_rename_dir() and sysfs_move_dir() grab all dentries using
sysfs_get_dentry() on startup.
* Dentries for all shadowed directories are pinned in memory to serve
as lookup start point.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Some sysfs operations require dentry and inode. sysfs_get_dentry()
looks up and gets dentry for the specified sysfs_dirent. It finds the
first ancestor with dentry attached and starts looking up dentries
from there.
Looking up from the nearest ancestor is necessary to support shadowed
directories because we can't reliably lookup dentry for one of the
shadows. Dentries for each shadow will be pinned in memory such that
they can serve as the starting point for dentry lookup.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
After add/remove path restructuring, the only user of
sysfs_drop_dentry() is sysfs_addrm_finish(). Move sysfs_drop_dentry()
to dir.c and make it static.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The original add/remove code had the following problems.
* parent's timestamps are updated on dentry instantiation. this is
incorrect with reclaimable files.
* updating parent's timestamps isn't synchronized.
* parent nlink update assumes the inode is accessible which won't be
true once directory dentries are made reclaimable.
This patch restructures add/remove paths to resolve the above
problems. Add/removal are done in the following steps.
1. sysfs_addrm_start() : acquire locks including sysfs_mutex and other
resources.
2-a. sysfs_add_one() : add new sd. linking the new sd into the
children list is caller's responsibility.
2-b. sysfs_remove_one() : remove a sd. unlinking the sd from the
children list is caller's responsibility.
3. sysfs_addrm_finish() : release all resources and clean up.
Steps 2-a and/or 2-b can be repeated multiple times.
Parent's inode is looked up during sysfs_addrm_start(). If available
(always at the moment), it's pinned and nlink is updated as sd's are
added and removed. Timestamps are updated during finish if any sd has
been added or removed. If parent's inode is not available during
start, sysfs_mutex ensures that parent inode is not created till
add/remove is complete.
All the complexity is contained inside the helper functions.
Especially, dentry/inode handling is properly hidden from the rest of
sysfs which now mostly operate on sysfs_dirents. As an added bonus,
codes which use these helpers to add and remove sysfs_dirents are now
more structured and simpler.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
As kobj sysfs dentries and inodes are gonna be made reclaimable,
i_mutex can't be used to protect sysfs_dirent tree. Use sysfs_mutex
globally instead. As the whole tree is protected with sysfs_mutex,
there is no reason to keep sysfs_rename_sem. Drop it.
While at it, add docbook comments to functions which require
sysfs_mutex locking.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Replace sysfs_lock and kobj_sysfs_assoc_lock with sysfs_assoc_lock.
sysfs_lock was originally to be used to protect sysfs_dirent tree but
mutex seems better choice, so there is no reason to keep sysfs_lock
separate. Merge the two spinlocks into one.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
As kobj sysfs dentries and inodes are gonna be made reclaimable,
dentry can't be used as naming token for sysfs file/directory, replace
kobj->dentry with kobj->sd. The only external interface change is
shadow directory handling. All other changes are contained in kobj
and sysfs.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Implement sysfs_find_dirent() and sysfs_get_dirent().
sysfs_dirent_exist() is replaced by sysfs_find_dirent(). These will
be used to make directory entries reclamiable.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Implement SYSFS_FLAG_REMOVED flag which currently is used only to
improve sanity check in sysfs_deactivate(). The flag will be used to
make directory entries reclamiable.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Rename sysfs_dirent->s_type to s_flags, pack type into lower eight
bits and reserve the rest for flags. sysfs_type() can used to access
the type. All existing sd->s_type accesses are converted to use
sysfs_type(). While at it, type test is changed to equality test
instead of bit-and test where appropriate.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
sysfs_drop_dentry() used to go through sd->s_dentry and
sd->s_parent->s_dentry to access the inodes. This is incorrect
because inode can be cached without dentry.
This patch makes sysfs_drop_dentry() access inodes using ilookup() on
sd->s_ino. This is both correct and simpler.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Fix oops on x86_64 caused by the dereference of dir in
sysfs_drop_dentry() made before checking if dir is not NULL
(cf. http://marc.info/?l=linux-kernel&m=118151626704924&w=2).
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Make sysfs_dirent use singly linked list for its tree structure.
sysfs_link_sibling() and sysfs_unlink_sibling() functions are added to
handle simpler cases. It adds some complexity and cpu cycle overhead
but reduced memory footprint is worthwhile on big machines.
This change reduces the sizeof sysfs_dirent from 104 to 88 on 64bit
and from 60 to 52 on 32bit.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Make sysfs_dirent->s_active an atomic_t instead of rwsem. This
reduces the size of sysfs_dirent from 136 to 104 on 64bit and from 76
to 60 on 32bit with lock debugging turned off. With lock debugging
turned on the reduction is much larger.
s_active starts at zero and each active reference increments s_active.
Putting a reference decrements s_active. Deactivation subtracts
SD_DEACTIVATED_BIAS which is currently INT_MIN and assumed to be small
enough to make s_active negative. If s_active is negative,
sysfs_get() no longer grants new references. Deactivation succeeds
immediately if there is no active user; otherwise, it waits using a
completion for the last put.
Due to the removal of lockdep tricks, this change makes things less
trickier in release_sysfs_dirent(). As all the complexity is
contained in three s_active functions, I think it's more readable this
way.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
These functions are about to receive more complexity and doesn't
really need to be inlined in the first place. Move them from
fs/sysfs/sysfs.h to fs/sysfs/dir.c.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The root sysfs_dirent didn't point to the root dentry fix it.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
After dentry is reclaimed, sysfs always used to allocate new dentry
and inode if the file is accessed again. This causes problem with
operations which only pin the inode. For example, if inotify watch is
added to a sysfs file and the dentry for the file is reclaimed, the
next update event creates new dentry and new inode making the inotify
watch miss all the events from there on.
This patch fixes it by using iget_locked() instead of new_inode().
sysfs_new_inode() is renamed to sysfs_get_inode() and inode is
initialized iff the inode is newly allocated. sysfs_instantiate() is
responsible for unlocking new inodes.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Reorganize/clean up sysfs_new_inode() and sysfs_create().
* sysfs_init_inode() is separated out from sysfs_new_inode() and is
responsible for basic initialization.
* sysfs_instantiate() replaces the last step of sysfs_create() and is
responsible for dentry instantitaion.
* type-specific initialization is moved out to the callers.
* mode is specified only once when creating a sysfs_dirent.
* spurious list_del_init(&sd->s_sibling) dropped from create_dir()
This change is to
* prepare for inode allocation fix.
* separate alloc and init code for synchronization update.
* make dentry/inode initialization more flexible for later changes.
This patch doesn't introduce visible behavior change.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
sysfs_alloc_ino() isn't used out side of fs/sysfs/dir.c. Make it
static.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
sysfs is now completely out of driver/module lifetime game. After
deletion, a sysfs node doesn't access anything outside sysfs proper,
so there's no reason to hold onto the attribute owners. Note that
often the wrong modules were accounted for as owners leading to
accessing removed modules.
This patch kills now unnecessary attribute->owner. Note that with
this change, userland holding a sysfs node does not prevent the
backing module from being unloaded.
For more info regarding lifetime rule cleanup, please read the
following message.
http://article.gmane.org/gmane.linux.kernel/510293
(tweaked by Greg to not delete the field just yet, to make it easier to
merge things properly.)
Signed-off-by: Tejun Heo <htejun@gmail.com>
Cc: Cornelia Huck <cornelia.huck@de.ibm.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch reimplements sysfs_drop_dentry() such that remove_dir() can
use it to drop dentry instead of using a separate mechanism. With
this change, making directories reclaimable is much easier.
This patch used to contain fixes for two race conditions around
sd->s_dentry but that part has been separated out and included into
mainline early as commit 6aa054aadf and
dd14cbc994.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Consolidate sd <-> dentry association into sysfs_attach_dentry() and
call it after dentry and inode are properly set up. This is in
preparation of sysfs_drop_dentry() updates.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Now that sysfs_dirent can be disconnected from kobject on deletion,
there is no need to orphan each attribute files. All [bin_]attribute
nodes are automatically orphaned when the parent node is deleted.
Kill attribute file orphaning.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
sysfs: implement sysfs_dirent active reference and immediate disconnect
Opening a sysfs node references its associated kobject, so userland
can arbitrarily prolong lifetime of a kobject which complicates
lifetime rules in drivers. This patch implements active reference and
makes the association between kobject and sysfs immediately breakable.
Now each sysfs_dirent has two reference counts - s_count and s_active.
s_count is a regular reference count which guarantees that the
containing sysfs_dirent is accessible. As long as s_count reference
is held, all sysfs internal fields in sysfs_dirent are accessible
including s_parent and s_name.
The newly added s_active is active reference count. This is acquired
by invoking sysfs_get_active() and it's the caller's responsibility to
ensure sysfs_dirent itself is accessible (should be holding s_count
one way or the other). Dereferencing sysfs_dirent to access objects
out of sysfs proper requires active reference. This includes access
to the associated kobjects, attributes and ops.
The active references can be drained and denied by calling
sysfs_deactivate(). All active sysfs_dirents must be deactivated
after deletion but before the default reference is dropped. This
enables immediate disconnect of sysfs nodes. Once a sysfs_dirent is
deleted, it won't access any entity external to sysfs proper.
Because attr/bin_attr ops access both the node itself and its parent
for kobject, they need to hold active references to both.
sysfs_get/put_active_two() helpers are provided to help grabbing both
references. Parent's is acquired first and released last.
Unlike other operations, mmapped area lingers on after mmap() is
finished and the module implement implementing it and kobj need to
stay referenced till all the mapped pages are gone. This is
accomplished by holding one set of active references to the bin_attr
and its parent if there have been any mmap during lifetime of an
openfile. The references are dropped when the openfile is released.
This change makes sysfs lifetime rules independent from both kobject's
and module's. It not only fixes several race conditions caused by
sysfs not holding onto the proper module when referencing kobject, but
also helps fixing and simplifying lifetime management in driver model
and drivers by taking sysfs out of the equation.
Please read the following message for more info.
http://article.gmane.org/gmane.linux.kernel/510293
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Implement bin_buffer which contains a mutex and pointer to PAGE_SIZE
buffer to properly synchronize accesses to per-openfile buffer and
prepare for immediate-kobj-disconnect.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
sysfs symlink is implemented by referencing dentry and kobject from
sysfs_dirent - symlink entry references kobject, dentry is used to
walk the tree. This complicates object lifetimes rules and is
dangerous - for example, there is no way to tell to which module the
target of a symlink belongs and referencing that kobject can make it
linger after the module is gone.
This patch reimplements symlink using only sysfs_dirent tree. sd for
a symlink points and holds reference to the target sysfs_dirent and
all walking is done using sysfs_dirent tree. Simpler and safer.
Please read the following message for more info.
http://article.gmane.org/gmane.linux.kernel/510293
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
kobj->dentry can go away anytime unless the user controls when the
associated sysfs node is deleted. This patch implements
kobj_sysfs_assoc_lock which protects kobj->dentry. This will be used
to maintain kobj based API when converting sysfs to use sysfs_dirent
tree instead of dentry/kobject.
Note that this lock belongs to kobject/driver-model not sysfs. Once
sysfs is converted to not use kobject in its interface, this can be
removed from sysfs.
This is in preparation of object reference simplification.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Make sd->s_element a union of sysfs_elem_{dir|symlink|attr|bin_attr}
and rename it to s_elem. This is to achieve...
* some level of type checking : changing symlink to point to
sysfs_dirent instead of kobject is much safer and less painful now.
* easier / standardized dereferencing
* allow sysfs_elem_* to contain more than one entry
Where possible, pointer is obtained by directly deferencing from sd
instead of going through other entities. This reduces dependencies to
dentry, inode and kobject. to_attr() and to_bin_attr() are unused now
and removed.
This is in preparation of object reference simplification.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Add s_name to sysfs_dirent. This is to further reduce dependency to
the associated dentry. Name is copied for directories and symlinks
but not for attributes.
Where possible, name dereferences are converted to use sd->s_name.
sysfs_symlink->link_name and sysfs_get_name() are unused now and
removed.
This change allows symlink to be implemented using sysfs_dirent tree
proper, which is the last remaining dentry-dependent sysfs walk.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Add sysfs_dirent->s_parent. With this patch, each sd points to and
holds a reference to its parent. This allows walking sysfs tree
without referencing sd->s_dentry which can go away anytime if the user
doesn't control when it's deleted.
sd->s_parent is initialized and parent is referenced in
sysfs_attach_dirent(). Reference to parent is released when the sd is
released, so as long as reference to a sd is held, s_parent can be
followed.
dentry walk in sysfs_readdir() is convereted to s_parent walk.
This will be used to reimplement symlink such that it uses only
sysfs_dirent tree.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Currently there are four functions to create sysfs_dirent -
__sysfs_new_dirent(), sysfs_new_dirent(), __sysfs_make_dirent() and
sysfs_make_dirent(). Other than sysfs_make_dirent(), no function has
two users if calls to implement other functions are excluded.
This patch consolidates sysfs_dirent creation functions into the
following two.
* sysfs_new_dirent() : allocate and initialize
* sysfs_attach_dirent() : attach to sysfs_dirent hierarchy and/or
associate with dentry
This simplifies interface and gives callers more flexibility. This is
in preparation of object reference simplification.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Error handling in sysfs_rename_dir() was broken.
* When lookup_one_len() fails, 0 is returned.
* If parent inode check fails, returns with inode mutex and rename
rwsem held.
This patch fixes the above bugs and flattens error handling such that
it's more readable and easier to modify.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Flatten cleanup paths in sysfs_add_link() and create_dir() to improve
readability and ease further changes to these functions. This is in
preparation of object reference simplification.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Error handling in fs/sysfs/bin.c:write() was wrong because size_t
count is used to receive return value from flush_write() which is
negative on failure.
This patch updates write() such that int variable is used instead.
read() is updated the same way for consistency.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
sysfs used simple incrementing allocator which is not guaranteed to be
unique. This patch makes sysfs use ida to give each sd a unique and
packed inode number.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There is no reason this function should be inlined and soon to follow
sysfs object reference simplification will make it heavier. Move it
to dir.c.
Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Implement debugfs_rename() to allow renaming files/directories in debugfs.
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
I ran into a curious issue when a lock is being canceled. The
cancellation results in a lock request to the vfs layer instead of an
unlock request. This is particularly insidious when the process that
owns the lock is exiting. In that case, sometimes the erroneous lock is
applied AFTER the process has entered zombie state, preventing the lock
from ever being released. Eventually other processes block on the lock
causing a slow degredation of the system. In the 2.6.16 kernel this was
investigated on, the problem is compounded by the fact that the cl_sem
is held while blocking on the vfs lock, which results in most processes
accessing the nfs file system in question hanging.
In more detail, here is how the situation occurs:
first _nfs4_do_setlk():
static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int reclaim)
...
ret = nfs4_wait_for_completion_rpc_task(task);
if (ret == 0) {
...
} else
data->cancelled = 1;
then nfs4_lock_release():
static void nfs4_lock_release(void *calldata)
...
if (data->cancelled != 0) {
struct rpc_task *task;
task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
data->arg.lock_seqid);
The problem is the same file_lock that was passed in to _nfs4_do_setlk()
gets passed to nfs4_do_unlck() from nfs4_lock_release(). So the type is
still F_RDLCK or FWRLCK, not F_UNLCK. At some point, when cancelling the
lock, the type needs to be changed to F_UNLCK. It seemed easiest to do
that in nfs4_do_unlck(), but it could be done in nfs4_lock_release().
The concern I had with doing it there was if something still needed the
original file_lock, though it turns out the original file_lock still
needs to be modified by nfs4_do_unlck() because nfs4_do_unlck() uses the
original file_lock to pass to the vfs layer, and a copy of the original
file_lock for the RPC request.
It seems like the simplest solution is to force all situations where
nfs4_do_unlck() is being used to result in an unlock, so with that in
mind, I made the following change:
Signed-off-by: Frank Filz <ffilzlnx@us.ibm.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Use the destination address of the original NLM request as the
source address in callbacks to the client.
Signed-off-by: Frank van Maarseveen <frankvm@frankvm.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Consider the case where the user has mounted the remote filesystem
server:/foo on the two local directories /bar and /baz using the
nosharedcache mount option. The files /bar/file and /baz/file are
represented by different inodes in the local namespace, but refer to the
same file /foo/file on the server.
Consider the case where a process opens both /bar/file and /baz/file, then
closes /bar/file: because the nfs4_state is not shared between /bar/file
and /baz/file, the kernel will see that the nfs4_state for /bar/file is no
longer referenced, so it will send off a CLOSE rpc call. Unless the
open_owners differ, then that CLOSE call will invalidate the open state on
/baz/file too.
Conclusion: we cannot share open state owners between two different
non-shared mount instances of the same filesystem.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Unless the user sets the NFS_MOUNT_NOSHAREDCACHE mount flag, we should
return EBUSY if the filesystem is already mounted on a superblock that
has set conflicting mount options.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Prior to David Howell's mount changes in 2.6.18, users who mounted
different directories which happened to be from the same filesystem on the
server would get different super blocks, and hence could choose different
mount options. As long as there were no hard linked files that crossed from
one subtree to another, this was quite safe.
Post the changes, if the two directories are on the same filesystem (have
the same 'fsid'), they will share the same super block, and hence the same
mount options.
Add a flag to allow users to elect not to share the NFS super block with
another mount point, even if the fsids are the same. This will allow
users to set different mount options for the two different super blocks, as
was previously possible. It is still up to the user to ensure that there
are no cache coherency issues when doing this, however the default
behaviour will be to share super blocks whenever two paths result in
the same fsid.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Hook in final components required for supporting in-kernel mount option
parsing for NFSv2 and NFSv3 mounts.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
For NFSv2 and v3 mounts, the first step is to contact the server's MOUNTD
and request the file handle for the root of the mounted share. Add a
function to the NFS client that handles this operation.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
This generic infrastructure works for both NFS and NFSv4 mounts.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Clean up white space and coding conventions.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
In preparation for supporting NFSv2 and NFSv3 mount option handling in the
kernel NFS client, convert mount_clnt.c to be a permanent part of the NFS
client, instead of built only when CONFIG_ROOT_NFS is enabled.
In addition, we also replace the "struct sockaddr_in *" argument with
something more generic, to help support IPv6 at some later point.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
A couple of callers just use a stringified IP address for the rpc client's
hostname. Move the logic for constructing this into rpc_create(), so it can
be shared.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
In preparation for handling NFS mount option parsing in the kernel,
rename rpcb_getport_external as rpcb_get_port_sync, and make it available
always (instead of only when CONFIG_ROOT_NFS is enabled).
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Refactor NFSv4 mount processing to break out mount data validation
in the same way it's broken out in the NFSv2/v3 mount path.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Move error handling code out of the main code path. The switch statement
was also improperly indented, according to Documentation/CodingStyle. This
prepares nfs_validate_mount_data for the addition of option string parsing.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
NFS and NFSv4 mounts can now share server address sanity checking. And, it
provides an easy mechanism for adding IPv6 address checking at some later
point.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Cc: Aurelien Charbon <aurelien.charbon@ext.bull.net>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
/home/cel/linux/fs/nfs/super.c: In function 'nfs_pseudoflavour_to_name':
/home/cel/linux/fs/nfs/super.c:270: warning: comparison between signed and unsigned
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The error return logic in nfs_get_sb now matches nfs4_get_sb, and is more maintainable.
A subsequent patch will take advantage of this simplification.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The new string utility function strndup_user can be used instead of
nfs_copy_user_string, eliminating an unnecessary duplication of function.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
inode->i_blocks is a blkcnt_t these days, which can be a u64 or unsigned
long, depending on the setting of CONFIG_LSF.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Rely on nfs4_try_open_cached() when appropriate.
Also fix an RCU violation in _nfs4_do_open_reclaim()
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
If we already have a stateid with the correct open mode for a given file,
then we can reuse that stateid instead of re-issuing an OPEN call without
violating the close-to-open caching semantics.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
We should not be calling open() on an inode that has a delegation unless
we're doing a reclaim.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Currently we force a synchronous call to __nfs_revalidate_inode() in
nfs_inode_set_delegation(). This not only ensures that we cannot call
nfs_inode_set_delegation from an asynchronous context, but it also slows
down any call to open().
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
There appear to be some rogue servers out there that issue multiple
delegations with different stateids for the same file. Ensure that when we
return delegations, we do so on a per-stateid basis rather than a per-file
basis.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
This ensures that nfs4_open_release() and nfs4_open_confirm_release()
can now handle an eventual delegation that was returned with out open.
As such, it fixes a delegation "leak" when the user breaks out of an open
call.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The test for state->state == 0 does not tell you that the stateid is in the
process of being freed. It really tells you that the stateid is not yet
initialised...
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Currently we do not check for the FMODE_EXEC flag as we should. For that
particular case, we need to perform an ACCESS call to the server in order
to check that the file is executable.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The maximum size depends on the filename size and a number of other
elements which are currently not being counted.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
There is no justification for keeping a special spinlock for the exclusive
use of the NFS writeback code.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
We should almost always be deferencing the rpc_auth struct by means of the
credential's cr_auth field instead of the rpc_clnt->cl_auth anyway. Fix up
that historical mistake, and remove the macro that propagated it.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Replace it with explicit calls to rpc_shutdown_client() or
rpc_destroy_client() (for the case of asynchronous calls).
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Also ensure that nfs_inode ncommit and npages are large enough to represent
all possible values for the number of pages.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Also get rid of a redundant call to nfs_setattr_update_inode(). The call to
nfs3_proc_setattr() already takes care of that.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The Linux NFS4 client simply skips over the bitmask in an O_EXCL open
call and so it doesn't bother to reset any fields that may be holding
the verifier. This patch has us save the first two words of the bitmask
(which is all the current client has #defines for). The client then
later checks this bitmask and turns on the appropriate flags in the
sattr->ia_verify field for the following SETATTR call.
This patch only currently checks to see if the server used the atime
and mtime slots for the verifier (which is what the Linux server uses
for this). I'm not sure of what other fields the server could
reasonably use, but adding checks for others should be trivial.
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
nfs_symlink() allocates a GFP_KERNEL page for the pagecache. Most
pagecache pages are allocated using GFP_HIGHUSER, and there's no reason
not to do that in nfs_symlink() as well.
Signed-off-by: Jeff Layton <jlayton@redhat.com>
We don't need to revalidate the fsid on the root directory. It suffices to
revalidate it on the current directory.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
nfs4_do_close() does not currently have any way to ensure that the user
won't attempt to unmount the partition while the asynchronous RPC call
is completing. This again may cause Oopses in nfs_update_inode().
Add a vfsmount argument to nfs4_close_state to ensure that the partition
remains mounted while we're closing the file.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
A number of race conditions may currently ensue if the user presses ^C
and then unmounts the partition while an asynchronous open() is in
progress.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Since PG_uptodate may now end up getting set during the call to
nfs_wb_page(), we can avoid putting a read request on the wire in those
situations.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The write may fail, so we should not mark the page as uptodate until we are
certain that the data has been accepted and written to disk by the server.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
There is no need to fail the entire O_DIRECT read/write just because
get_user_pages() returned fewer pages than we requested.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Many places in kernel use seq_file API to iterate over a regular list_head.
The code for such iteration is identical in all the places, so it's worth
introducing a common helpers.
This makes code about 300 lines smaller:
The first version of this patch made the helper functions static inline
in the seq_file.h header. This patch moves them to the fs/seq_file.c as
Andrew proposed. The vmlinux .text section sizes are as follows:
2.6.22-rc1-mm1: 0x001794d5
with the previous version: 0x00179505
with this patch: 0x00179135
The config file used was make allnoconfig with the "y" inclusion of all
the possible options to make the files modified by the patch compile plus
drivers I have on the test node.
This patch:
Many places in kernel use seq_file API to iterate over a regular list_head.
The code for such iteration is identical in all the places, so it's worth
introducing a common helpers.
Signed-off-by: Pavel Emelianov <xemul@openvz.org>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We re-use the RESVSP/UNRESVSP ioctls from xfs which allow the user to
allocate and deallocate regions to a file without zeroing data or changing
i_size.
Though renamed, the structure passed in from user is identical to struct
xfs_flock64. The three fields that are actually used right now are l_whence,
l_start and l_len.
This should get ocfs2 immediate compatibility with userspace software using
the pre-existing xfs ioctls.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Provide an internal interface for the removal of arbitrary file regions.
ocfs2_remove_inode_range() takes a byte range within a file and will remove
existing extents within that range. Partial clusters will be zeroed so that
any read from within the region will return zeros.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
The partial cluster zeroing code used during truncate usually assumes that
the rightmost byte in the range to be zeroed lies on a cluster boundary.
This makes sense for truncate, but punching holes might require zeroing on
non-aligned rightmost boundaries.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Add code to the btree paths to support the removal of arbitrary regions
within an existing extent. With proper higher level support this can be used
to "punch holes" in a file. Truncate (a special case of hole punching) could
also be converted to use these methods.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
This can now be trivially supported with re-use of our existing extend code.
ocfs2_allocate_unwritten_extents() takes a start offset and a byte length
and iterates over the inode, adding extents (marked as unwritten) until len
is reached. Existing extents are skipped over.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Update the write code to detect when the user is asking to write to an
unwritten extent. Like writing to a hole, we must zero the region between
the write and the cluster boundaries. Most of the existing cluster zeroing
logic can be re-used with some additional checks for the unwritten flag on
extent records.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Writes to a region marked as unwritten might result in a record split or
merge. We can support splits by making minor changes to the existing insert
code. Merges require left rotations which mostly re-use right rotation
support functions.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
The top level calls and logic for growing a tree can easily be abstracted
out of ocfs2_insert_extent() into a seperate function - ocfs2_grow_tree().
This allows future code to easily grow btrees when needed.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Now that we have a method to deallocate blocks from them, each node should
allocate extent blocks from their local suballocator file.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Deallocation of suballocator blocks, most notably extent blocks, might
involve multiple suballocator inodes.
The locking for this can get extremely complicated, especially when the
suballocator inodes to delete from aren't known until deep within an
unrelated codepath.
Implement a simple scheme for recording the blocks to be unlinked so that
the actual deallocation can be done in a context which won't deadlock.
Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
